Exploratory Data Analysis

Exploration

Missing Data

Scaled and Combined

The idea behind this model is that teams that are better than the average will win more games and teams worse than the average will win less. The way we determine if a team is better than average is by looking at how well they preform at batting, pitching, and fielding.

Since there are more than one way to win a baseball game (i.e. have some power sluggers that hit home runs, vs have really good single batters.) we need to combine the various batting measures. Now since getting a strikout at bat is bad, we need to change the sign of this variable. That way it can be combined and will fit the better teams win more and worse teams less model.

We are going to scale all variables . That centers them at 0 and gives them a standard deviation of 1. We can then combine almost all the batting variables into one measure (hit by pitcher is excluded).

Call:
lm(formula = TARGET_WINS ~ TEAM_BATTING + TEAM_PITCHING + TEAM_FIELDING, 
    data = training)

Residuals:
    Min      1Q  Median      3Q     Max 
-42.578  -8.197   0.281   8.526  49.155 

Coefficients:
              Estimate Std. Error t value Pr(>|t|)    
(Intercept)    80.8822     0.2885  280.33   <2e-16 ***
TEAM_BATTING    3.6415     0.1897   19.20   <2e-16 ***
TEAM_PITCHING   3.0643     0.2174   14.10   <2e-16 ***
TEAM_FIELDING  -0.1567     0.2062   -0.76    0.447    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 12.35 on 1986 degrees of freedom
  (286 observations deleted due to missingness)
Multiple R-squared:  0.2169,    Adjusted R-squared:  0.2157 
F-statistic: 183.3 on 3 and 1986 DF,  p-value: < 2.2e-16

This model says that the average baseball team will win about 81 games. If their batting is one standard deviation better than the average they will win 4 more games. They will win 3 more games for if their pitching is better than average and 0 if their fielding.